EROFS Adds Zstd & Btrfs Gets Minor Performance Work In Linux 6.10
The EROFS and Btrfs file-systems saw their feature patches merged as part of the ongoing Linux 6.10 merge window.
The read-only EROFS file-system this cycle saw improvements to its LZ4 code to better operate on large servers with hundreds of CPU cores. But the main new feature this cycle is Zstandard (Zstd) being available as an alternative compression algorithm for this file-system that is commonly used by containers and embedded/mobile hardware. The EROFS Zstd support has shown to perform better than LZ4 but at higher CPU overhead. The Zstd support for now is considered experimental.
All the EROFS feature changes for Linux 6.10 can be found via this pull request.
David Sterba with SUSE meanwhile sent out the Btrfs updates for Linux 6.10. This cycle there are some "minor" performance optimizations plus code refactoring and other low-level improvements. The Btrfs performance work includes inlining B-tree locking functions to help with metadata-heavy changes, relaxed locking on a range being reflinked, speeding up NOCOW write checks, and reducing extent locking ranges in various places.
Btrfs also saw more folio conversion work, extent locking optimizations, improved error handling, and other changes for Linux 6.10.
The read-only EROFS file-system this cycle saw improvements to its LZ4 code to better operate on large servers with hundreds of CPU cores. But the main new feature this cycle is Zstandard (Zstd) being available as an alternative compression algorithm for this file-system that is commonly used by containers and embedded/mobile hardware. The EROFS Zstd support has shown to perform better than LZ4 but at higher CPU overhead. The Zstd support for now is considered experimental.
All the EROFS feature changes for Linux 6.10 can be found via this pull request.
David Sterba with SUSE meanwhile sent out the Btrfs updates for Linux 6.10. This cycle there are some "minor" performance optimizations plus code refactoring and other low-level improvements. The Btrfs performance work includes inlining B-tree locking functions to help with metadata-heavy changes, relaxed locking on a range being reflinked, speeding up NOCOW write checks, and reducing extent locking ranges in various places.
Btrfs also saw more folio conversion work, extent locking optimizations, improved error handling, and other changes for Linux 6.10.
2 Comments